Lucebox
A dedicated workstation for running large open AI models locally without per-token fees.
The Lucebox Zero 495 pairs an AMD Ryzen AI MAX+ PRO CPU with a 32 GB Radeon AI PRO R9700 GPU and up to 192 GB of unified memory, so models too big for the GPU's own VRAM still run using the shared pool. It ships with custom kernels and speculative decoding tuned for this hardware, and setup is plug-and-play from lucebox.com. At full inference load it draws about 500W, idling around 40W, backed by a 1-year parts-and-labor warranty after 72 hours of burn-in.
What holds up
- +Up to 192 GB unified memory lets oversized models run via the shared CPU/GPU pool instead of failing to load
- +No per-token billing — one hardware purchase replaces ongoing inference fees
- +Optional cluster kit (dual 100 GbE) lets you link multiple units for bigger workloads
Mind the limits
- −11.97 L chassis draws up to 500W under load — not a low-power or silent option
- −Backed by Y Combinator funding but no price is listed on the page itself, so cost isn't upfront
Picked here? Add the badge
Put it on your site. It links back to this page. Free for every listed product.
<a href="https://stillworks.dev/products/p/lucebox/"><img src="https://stillworks.dev/badge/lucebox.svg" alt="Picked by StillWorks" width="230" height="54"></a>